Blind Denoising Research Articles

Blind Image Restoration Based on Cycle-Consistent Network

This paper studies the blind image restoration where the ground truth is unavailable and the downsampling process is unknown. This complicated setting makes supervised learning and accurate kernel estimation impossible. Inspired by the recent success of image-to-image translation, this paper resorts to the unsupervised Cycle-consistent based framework to tackle this challenging problem. Different from the image-to-image task, the fidelity of reconstructed image is important for image restoration. Therefore, to improve the reconstruction ability of the Cycle-consistent network, we make explorations from the following aspects. First, we constrain low-frequency content in data to preserve the content of output from LR input. Second, we impose constraint on the content of training data to provide better supervision for discriminator, helping to suppress high-frequency artifacts or fake textures. Third, we average model parameters to further improve the generated image quality and help with model selection for GAN-based methods. Since GAN-based methods tend to produce various artifacts with different models, model average could realize a smoother control of balancing artifacts and fidelity. We have conducted extensive experiments on real noise and super resolution datasets to validate the effectiveness of the above techniques. The proposed ECycleGAN also demonstrates superior performance to SOTA methods in two applications -- blind SR and blind denoising.

A dual of real image noise-related blind denoising technique

A dual of real image noise-related blind denoising technique

Dynamic chaotic Gold-Panning Optimizer and its typical engineering applications

Swarm intelligence algorithms are one of the key technologies in solving optimization problems for practical engineering applications, such as mechanical structure design, image analysis, process flow design, etc. Higher accuracy and efficiency mean better comprehensive performance in the practical engineering system based on optimization methods. Gold-Panning Algorithm is one of the swarm intelligence algorithms proposed in 2021, for solving image segmentation problems. However, the self decision-making mechanism based on multi-agent information interaction introduced in it weakens its convergence ability, resulting in its ability to exploit potential solutions being limited. Hence, chaotic maps were introduced to improve the optimization capacity and efficiency, which can provide a more reliable and effective ability to explore and exploit potential optimal solutions in different iteration stages. Moreover, a dynamic selection strategy is utilized to choose the better step-size iteration scheme between the Gaussian distribution and Levy flight. It can further strengthen the exploitation capability of the Gold-Panning Algorithm, reducing the possibility of premature convergence. Based on CEC’2020 benchmark functions, Dynamic Chaotic Gold-Panning Optimizer is compared with the other meta-heuristic algorithms to evaluate its performance and the results shows strong competitiveness in both robustness and accuracy for solving optimization problems. Then, based on the proposed optimizer, its binary variant is applied in feature selection. Besides, combined with the bilateral filtering and threshold segmentation model, the image blind denoising and image segmentation optimization schemes are proposed respectively. Simulation results indicate it presents an excellent comprehensive performance in solving the corresponding engineering task.

Noise2Kernel: Adaptive Self-Supervised Blind Denoising Using a Dilated Convolutional Kernel Architecture.

With the advent of unsupervised learning, efficient training of a deep network for image denoising without pairs of noisy and clean images has become feasible. Most current unsupervised denoising methods are built on self-supervised loss with the assumption of zero-mean noise under the signal-independent condition, which causes brightness-shifting artifacts on unconventional noise statistics (i.e., different from commonly used noise models). Moreover, most blind denoising methods require a random masking scheme for training to ensure the invariance of the denoising process. In this study, we propose a dilated convolutional network that satisfies an invariant property, allowing efficient kernel-based training without random masking. We also propose an adaptive self-supervision loss to increase the tolerance for unconventional noise, which is specifically effective in removing salt-and-pepper or hybrid noise where prior knowledge of noise statistics is not readily available. We demonstrate the efficacy of the proposed method by comparing it with state-of-the-art denoising methods using various examples.

Deep learning optical image denoising research based on principal component estimation.

High-quality denoising of optical interference images usually requires preliminary prediction of the noise level. Although blind denoising can filter the image at the pixel level without noise prediction, it inevitably loses a significant amount of phase information. This paper proposes a fast and high-quality denoising algorithm for optical interference images that combines the merits of a principal component analysis (PCA) and residual neural networks. The PCA is used to analyze the image noise and, in turn, establishes an accurate mapping between the estimated and true noise levels. The mapping helps to select a suitable residual neural network model for image processing, which maximizes the retention of image information and reduces the effect of noise. In addition, a comprehensive evaluation factor to account for the time complexity and denoising effect of the algorithm is proposed, since time complexity can be a dominant concern in some cases of actual measurement. The performance of the denoising algorithm and the effectiveness of the evaluation criterion are demonstrated to be high by processing a set of optical interference images and benchmarking with other denoising algorithms. The proposed algorithm outperforms the previously reported counterparts in a specific area of optical interference image preprocessing and provides an alternative paradigm for other denoising problems of optics, such as holograms and structured light measurements.

Blind denoising using dense hybrid convolutional network

The performance of existing deep convolutional networks is limited when encountering images with different noise levels. In this study, a denoising method with state-of-the-art performance that combines a deep convolutional network with the traditional nonlocal mean denoising method is proposed. The noisy image is first denoised using the nonlocal mean method. Then, the denoised image is input into the proposed dense hybrid convolutional network to be trained, producing a clean image with clear details. The dense hybrid convolutional network comprises three parts: a feature-extracting noise-suppressing module that extracts abstract features from denoised images and suppresses the residual noise by interval convolution; a feature-learning module used for training blurred edges and textures; and a magnifying module that uses deconvolution to restore the feature maps to the original size and reduce the noise again. In contrast to existing denoising algorithms, the method has two desirable properties: 1) it can restore edges and textures clearly while removing the noise; 2) it effectively deals with noise of unknown levels (i.e. blind denoising) with a single network model. The conducted experiments show that the proposed method achieves superior performance compared to those of state-of-the-art denoising methods.

Multiple Residual Dense Networks for Reconfigurable Intelligent Surfaces Cascaded Channel Estimation

Reconfigurable intelligent surface (RIS) constitutes an essential and promising paradigm that relies programmable wireless environment and provides capability for space-intensive communications, due to the use of low-cost massive reflecting elements over the entire surfaces of man-made structures. However, accurate channel estimation is a fundamental technical prerequisite to achieve the huge performance gains from RIS. By leveraging the low rank structure of RIS channels, three practical residual neural networks, named convolutional blind denoising network, convolutional denoising generative adversarial networks and multiple residual dense network, are proposed to obtain accurate channel state information, which can reflect the impact of different methods on the estimation performance. Simulation results reveal the evolution direction of these three methods and reveal their superior performance compared with existing benchmark schemes.

Interferogram blind denoising using deep residual learning for phase-shifting interferometry

The interferogram containing the noises often affects the accuracy of phase retrieval, leading to the degradation of the phase imaging quality. To address this issue, a new interferogram blind denoising (IBD) method based on deep residual learning is proposed. In the presence of unknown noise levels, during the training, the deep residual convolutional neural networks (DRCNN) in the IBD approach is able to remove the latent clean interferogram implicitly, and then gradually establish the residual mapping relation in the pixel-level between the interferogram and the noises. With a well-trained DRCNN model, this algorithm can deal not only with the single-frame interferogram efficiently but also with the multi-frame phase-shifted interferograms collaboratively, while effectively retaining interferogram features related to phase retrieval. Simulation and experimental results demonstrate the feasibility and applicability of the proposed IBD method.

CDLNet: Noise-Adaptive Convolutional Dictionary Learning Network for Blind Denoising and Demosaicing

Deep learning based methods hold state-of-the-art results in low-level image processing tasks, but remain difficult to interpret due to their black-box construction. Unrolled optimization networks present an interpretable alternative to constructing deep neural networks by deriving their architecture from classical iterative optimization methods without use of tricks from the standard deep learning tool-box. So far, such methods have demonstrated performance close to that of state-of-the-art models while using their interpretable construction to achieve a comparably low learned parameter count. In this work, we propose an unrolled convolutional dictionary learning network (CDLNet) and demonstrate its competitive denoising and joint denoising and demosaicing (JDD) performance both in low and high parameter count regimes. Specifically, we show that the proposed model outperforms state-of-the-art fully convolutional denoising and JDD models when scaled to a similar parameter count. In addition, we leverage the model's interpretable construction to propose a noise-adaptive parameterization of thresholds in the network that enables state-of-the-art blind denoising performance, and near-perfect generalization on noise-levels unseen during training. Furthermore, we show that such performance extends to the JDD task and unsupervised learning.

Gradient and Multi Scale Feature Inspired Deep Blind Gaussian Denoiser

In this paper, a novel deep blind Gaussian denoising network is proposed utilizing the concepts of gradient information, multi-scale feature information and feature denoising for removing additive white Gaussian noise (AWGN) from images. The proposed network consists of two modules where in the first module generates an intermediate image whose gradient information is concatenated with the features of second module to generate the final residual image. Subtracting this residual image with the noisy image gives the desired denoised image. The feature denoising block used in the middle of the first module enhances the feature information of the intermediate image. The usage of gradient information of this intermediate denoised image, together with the multi scale feature information block, in the second module further contributes to the quality of the final denoised image. Experimental results show superior denoising performance of the proposed method in comparison to several state of the art classical and learning based blind denoising methods like EPLL, BM3D, WNNM, DnCNN, MemNet, BUIFD, Self2Self and ComplexNet by a decent margin (an improvement of up to 2.4dB in PSNR, 0.07 in SSIM and 0.03 in FOM index with the second best performing model) when experimented over BSD68, Set5, Set14, SunHays80 and Manga109 image databases.

ISCL: Interdependent Self-Cooperative Learning for Unpaired Image Denoising.

With the advent of advances in self-supervised learning, paired clean-noisy data are no longer required in deep learning-based image denoising. However, existing blind denoising methods still require the assumption with regard to noise characteristics, such as zero-mean noise distribution and pixel-wise noise-signal independence; this hinders wide adaptation of the method in the medical domain. On the other hand, unpaired learning can overcome limitations related to the assumption on noise characteristics, which makes it more feasible for collecting the training data in real-world scenarios. In this paper, we propose a novel image denoising scheme, Interdependent Self-Cooperative Learning (ISCL), that leverages unpaired learning by combining cyclic adversarial learning with self-supervised residual learning. Unlike the existing unpaired image denoising methods relying on matching data distributions in different domains, the two architectures in ISCL, designed for different tasks, complement each other and boost the learning process. To assess the performance of the proposed method, we conducted extensive experiments in various biomedical image degradation scenarios, such as noise caused by physical characteristics of electron microscopy (EM) devices (film and charging noise), and structural noise found in low-dose computer tomography (CT). We demonstrate that the image quality of our method is superior to conventional and current state-of-the-art deep learning-based unpaired image denoising methods.

Remote sensing image super‐resolution based on convolutional blind denoising adaptive dense connection

The current super-resolution (SR) deep network is mainly applied to the common image and pays little attention to the image with noise. The remote sensing image contains much noise, so that the SR reconstruction effect is not satisfactory. Therefore, a convolution blind denoising adaptive dense connection SR (CBD-ADCSR) network for the remote sensing image is proposed in this paper. The whole model is divided into a convolution blind denoising (CBD) network for denoising and an ADCSR network for reconstruction. Firstly, the components of the network are given in detail and are analysed. Secondly, a data set making method is designed combining motion blur, defocusing blur and Gaussian noise, which is used to generate low-resolution image data sets with complex degradation for the model training. Finally, through the detailed comparative experiment, it is proved that the reconstruction effect of the CBD-ADCSR model is better than that of the most state-of-the-art algorithms in objective criteria. In addition, compared with the original ADCSR network, CBD-ADCSR has a stronger ability for noise suppression.

Dynamic coarse‐to‐fine ISAR image blind denoising using active joint prior learning

Most existing nonblind denoising approaches assumed the noise to be homogeneous white Gaussian distribution with known intensity. However, it is difficult to know beforehand or model accurately real-world noises with complex hybrid distribution and noise intensity. In this paper, active joint prior learning (JPL) is proposed for real-world ISAR image blind denoising. (1) To explore strong model hierarchy and components relationship automatically, a novel graphical Dirichlet mixture process (GDMP) model is developed, where the latent representations and component hyperparameters are jointly learned from each other. (2) A multiscale joint learning strategy (MJLS) is proposed to take advantage of both the optimization- and discriminative learning-based capabilities. The external noiseless, internal noisy image information and their relationships are jointly explored simultaneously. (3) Low-rank weighted sparse learning (LWSL) is proposed to learn sparse discriminative correlation components for robust prior learning, and latent low-rank embedding for GDMP patterns self-adaptive inference. Extensive experimental results on ISAR image datasets demonstrate the effectiveness of the proposed model for both synthesis and real-world noisy ISAR images, and the proposed method outperforms the state-of-the-art denoising methods.

Partial-DNet: A Novel Blind Denoising Model With Noise Intensity Estimation for HSI

Because of the inevitable noise interference in hyperspectral images (HSIs), the understanding and application of HSIs are seriously restricted. To solve this problem, the research on the data-driven neural network-based denoising method has become a hotspot in recent years. However, for HSIs with inconsistent and mixed noises, the traditional data-driven denoising algorithms have obvious limitations on generalization ability. In order to overcome these drawbacks, a novel partial densenet (Partial-DNet) model with noise intensity estimation is proposed for HSIs blind denoising in this article. In the proposed Partial-DNet model, the noise intensity of each band is estimated in the first and then fused with the observed images to generate the feature maps by introducing the channel attention mechanism. Finally, a novel multiscale neural network is explored to extract the spatial–spectral joint features. The contributions of this article can be summarized as follows: 1) the noise intensity of each band is estimated as a prior to guide the blind denoising framework suit with different data sets adaptively; 2) a Partial-DNet model is proposed to extract multiscale spatial–spectral features more efficient to maintain the details better while denoising; and 3) the experiments on both simulated and real HSI data sets indicate that the proposed denoising framework can be used to remove the inconsistent mixed noises in HSIs adaptively. Compared with the other state-of-the-art denoising methods, HSIs denoised by the proposed Partial-DNet algorithm not only have a higher PSNR index but also have higher classification accuracy under the same situation.

Robust dense light field reconstruction from sparse noisy sampling

Reconstructing densely-sampled light fields has attracted extensive attention recently. Numerous methods have been developed on the basis of various geometric characteristics of light field. However, these methods lose sight of the potential impact of input noise. Noise is ubiquitous in light field imaging of real-world scenes. The noise presented in the input views can propagate into the subsequent stages, and can significantly degrade the performance of the light field reconstruction. To remedy this problem, we propose a unified learning framework that learns the noise-invariant representations of light field, and reconstructs a clean densely-sampled light field from sparse noisy sampling. The novelty of this paper is the scheme to suppress input noise propagation and to improve the noise tolerance of view synthesis. Three convolutional sub-networks are jointly designed for this task: an encoder-decoder based unsupervised learning for the noise-independent estimation of light field depth, a residual learning based noise filtering for the unsupervised blind denoising of the sparse noisy views, and a complementary learning based rendering for the clean virtual view synthesis. We have made a comprehensive evaluation on 5 public datasets under different types of noise. The visual and quantitative comparisons demonstrate that our method not only exhibits significant superiority in noise tolerance, but also achieves excellent light field reconstruction performance.

Self-Supervised Learning for Seismic Data Reconstruction and Denoising

With their powerful feature extraction ability, convolutional neural network (CNN) models achieve excellent signal reconstruction and recovery performances compared with those of traditional methods. The CNN-based approaches mainly use supervised learning approaches; thus, they require large numbers of ground-truth labeled samples. However, in the seismic denoising field, collecting large numbers of labeled samples is impossible; thus, the main challenge to using deep learning methods is a lack of labeled data. Moreover, the data that are available contain noise. To resolve these shortcomings, this letter proposes a novel self-supervised learning framework to reconstruct and perform blind denoising of seismic data images; this approach requires no labeled training data. We utilize a masking procedure to modify an observation input to a CNN to create a <inline-formula xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink"> <tex-math notation="LaTeX">$\mathcal {J}$ </tex-math></inline-formula> -invariant function and incorporate a specific CNN architecture known as U <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> Net, which implements a two-level nested autoencoder that extracts complex feature information from different scales. We modify the network to make it more suitable for seismic signal reconstruction. Finally, we use the self-supervised loss between the original observation and the net output to update the weights of U <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> Net through backpropagation. Tests on both synthetic and field data demonstrate the superior performance of our algorithm on low signal-to-noise ratio data.

Hybrid BM3D and PDE filtering for non-parametric single image denoising

The BM3D method achieves excellent denoising performance, but it has artificial effects and bias effects and its performance largely depends on the noise level parameter. To address this, we propose a hybrid BM3D and PDE method for non-parametric single image denoising. First, a non-local Perona–Malik (NLPM) filtering is proposed, and we prove its discontinuity maintaining, mean invariance, convergence, and local continuity. Based on these mathematical properties, an NLPM based noise level estimator (NLPM-NLE) is explored, which involves three steps: preprocessing by NLPM filtering, sample area selection, parameter estimation. And then, we advance a stable-BM3D (SBM3D) method with NLPM filtering to avoid artificial effects and bias effects. Finally, connecting the NLPM-NLE and SBM3D by merging the same part, we develop a non-parametric single image denoising (NPSID) method. Additionally, our proposed BM3D method with NLPM-NLE and the NPSID are compared with other blind denoising methods including PCA + BM3D, WTP + BM3D, and ESM + BM3D on real image denoising. Experiments show that the proposed non-parametric method can automatically and effectively remove noise and preserve details.

Ultrasonic Logging Image Denoising Based on CNN and Feature Attention

Various kinds of noise will be produced during the process of ultrasonic logging in high temperature and high-pressure environment under oil wells, which is blurring the logging image. This paper presents a novel end-to-end denoising model (ULNet) based on CNN and feature attention to address this problem and remove the noise from ultrasonic logging images. Our method mainly includes feature attention, feature enhancement based on residual model and reconstruction for ultrasonic logging image. Feature enhancement based on a residual model integrates global and local features to increase the expressive ability of the denoising model. Feature attention is used to distinguish the channel feature weights, and effective for blind denoising of actual images. Kernel dilation and skip-connection is used to reduce the computational cost during training. The Noise mapping results are used to reconstruct a clean image. Comprehensive quantitative and qualitative evaluations of results for selected study datasets collected at six oil wells in China show that this model is a feasible and effective means for denoising ultrasonic logging images. Overall, ULNet shows potential for practical ultrasonic logging images denoising.

Denoising real bursts with squeeze‐and‐excitation residual network

The goal of image denoising is to recover a clean image from noisy input(s). For single image denoising, utilising similarities (or priors) within and across an image dataset helps recover clean images. As the noise level increases, using multiple frames become feasible, which is defined as burst denoising. In this study, the authors propose a deep residual model with squeeze-and-excitation (SE) modules for the burst denoising. Unlike previous methods, the authors' model does not need an explicit aligning procedure, which is light-weighted and fast. The network contains a noise estimation convolutional neural network, which makes it capable of blind denoising. Besides, by inverting the image processing pipeline and simulating real noise in bursts, their model can suppress real noise blindly. Since denoising performance is closely related to the noise level, frame displacement, and the number of frames (burst length), intensive experiments including ablation study are performed. Quantitative results show that the proposed method performs significantly better than previous state-of-the-art methods V-BM4D and KPN in removing Gaussian noise. Qualitative results show that the proposed method is also effective in removing real noise using bursts and the SE module is key to reduce blur in results.

Multinoise-type blind denoising using a single uniform deep convolutional neural network

Deep convolutional neural networks (CNNs) have achieved considerable success with image denoising. However, they still lack consistent performance across different noise types and levels. We extend noise scenarios to four categories: Gaussian, random-impulse, salt-and-pepper, and Poisson. We also propose a multinoise-type blind denoising network (MBDNet) that solves the blind denoising task using a uniform deep CNN architecture. The network can be divided into two stages where a concise CNN is first used to estimate auxiliary noise-type and noise-level information. Estimation results are then integrated as additional channels of the noisy image and are fed to the subsequent denoising stage. A unique two-branch structure is further adopted in the residual denoising CNN, wherein a shallow branch predicts the filter-flow mask and adaptively adjusts the feature extraction of the parallel deep branch. Extensive experiments on synthetic noisy images validate the effectiveness of the noise-estimation and denoising subnetworks and show that MBDNet is highly competitive as compared to state-of-the-art methods in both denoising performance and model runtime.